Tony made an interesting point about the inclusion of SD-WAN within Gartner’s Magic Quadrant for WAN optimization controllers. It was a point that could help unlock overlooked expenditures in IT budgets.
“We believe WAN requirements are evolving rapidly as enterprises become more frustrated with the high cost and complexity of MPLS networking. There just hasn’t been an ability, until now, to easily leverage lower-cost Internet in a secure, controlled and optimized manner.”
But how much can you really save with an SD-WAN? Well, consider this: on average, IT spent 9 percent of its budget on the data network last year, and more than three-quarters of that expenditure went to transmission and personnel. SD-WAN addresses both problems.
Internet bandwidth continues to be a fraction of MPLS costs. A 10 Mbps MPLS port in New York costs $608 per month; a 150 Mbps FiOS connection runs $160. To put that another way, FiOS is about $1 per megabit; MPLS is just over $60 per megabit. Network engineers have been limited in taking advantage of those economics because of the availability, security, and performance issues of the Internet — issues addressed by Silver Peak.
But we all know there’s more to costing out a network than bandwidth. SD-WAN requires redundancy in the local access to match (or exceed) the uptime on the Internet. The SD-WAN software must be deployed at each location. Hardware may be needed for those nodes, though many organizations have extra CPU cycles they can leverage for SD-WANs. Engineering time is required to plan and deploy the new software. If you’re switching from an MPLS then contractual issues might kick in and as well.
So a 98 percent savings isn’t going to happen, but the cost benefits can still be substantial. A back-of-the-envelope calculation I recently ran involving one data center (10 Mbps) and nine remote offices (1.5 Mbps) with all software and hardware included still showed a return within year one, and then a 64% monthly savings starting in year two. Not quite 98 percent, but still substantial.
And what about those personnel costs? Nick Lippis in his introductory talk at the ONUG meeting last fall made the point that a single server engineer can manage 10,000 virtual machines (VMs), but a single network engineer can manage 400 devices. The policy-based automation and centralized managed tools we’ve seen in the virtual space have radically redefined the costs of server management.
Switching from MPLS to the Internet isn’t new, but with SD-WAN management the efficiencies are significantly improved. Business and policies inform how applications use the network. Auto-configuration allows SD-WAN nodes to be deployed in minutes — not days. We’ve seen how our own customers, such as at Perkins Coie, have been able to dramatically reduce deployment and upgrade times with an SD-WAN.
So some savings will be realized on efficiencies and even more with lower bandwidth costs. Now the only question is, what can your team create with five percent more capital?